almeidasrenato

codex-image-auto

Generates an image with the user's ChatGPT Plus (no API key) without asking anything - Claude picks the quota (Codex or GPT), model, effort and size itself, always prioritizing image quality. Use when the user runs /codex-image-auto, says to decide for them ("escolha você", "decide você", "sem perguntar"), or wants a quick image shown in the chat. For images that belong to the project being worked on (site/app assets), use codex-image-project instead.

codex-image-project

Generates images for the project currently being worked on, using the user's ChatGPT Plus (no API key), and saves them inside the project's own asset folder with descriptive names. Claude reads the project to match its folder conventions, visual identity and the size each image needs, and decides quota/model/effort prioritizing quality. Use when the user runs /codex-image-project, or asks for images for this project/site/app (hero, banners, product shots, illustrations, placeholders to replace).

codex-image

Generates an image with the user's ChatGPT Plus (no API key), asking which quota (Codex or GPT), model and effort to use. Codex = codex exec + $imagegen (gpt-image-2); GPT = chrome-use driving chatgpt.com in the user's logged-in Chrome. Use when the user runs /codex-image or asks to generate an image and wants to choose the options. If the user wants Claude to decide, use codex-image-auto instead.

codex-update-models

Updates the Codex and ChatGPT (GPT) model and reasoning-effort lists used by the codex-image skill. Codex is read from a local file; GPT briefly opens chatgpt.com in a tab of the user's logged-in Chrome to read the model menu (no message sent, no quota used). Use when the user runs /codex-update-models or asks to update or see the latest models.